Confirmed template parsing bug: bug in error message generation.

[Paul Derbyshire]

At 04:59 AM 3/1/99 -0300, you wrote:
>Then please post a bug report...

I already did!

>...containing a single code snippet that demonstrates the problem...

"extern foo<int> bar;"

Parse error, but line is well-formed; should be undeclared identifier
"foo". Caused a 4 hour work stoppage here with people trying to track down
obscure aspects of the standard regarding templates thinking they made a
subtly ill-formed template declaration, and then throwing "typename"
qualifiers around at random half-heartedly, before suspecting it might be a
compiler bug, all of it triggered by a simple use of a symbol not imported
from its namespace. :-)

>...and a description of the host in which you have observed the
>problem...

x86-MSDOS. Which anyone who has read the original stuff will know. Not that
there is a snowball's chance in hell that the target has any influence
whatsoever on the *parser* :-)

>otherwise lots of people may have to spend time re-creating the code 
>snippet you already have.

Why you believe this is beyond me, since the code snippet (above) was in
the very post you quoted IIRC, and the target, even if relevant for parser
issues, has been referred to be my in the past also.

Copied to egcs-bugs.

-- 
   .*.  "Clouds are not spheres, mountains are not cones, coastlines are not
-()  <  circles, and bark is not smooth, nor does lightning travel in a
   `*'  straight line."    -------------------------------------------------
        -- B. Mandelbrot  | http://surf.to/pgd.net
_____________________ ____|________     Paul Derbyshire     pderbysh@usa.net
Programmer & Humanist|ICQ: 10423848|